Obstructive Sleep Apnea and the Risk of Gout

A Population-Based Case-Control Study

Caroline van Durme; Bart Spaetgens; Johanna Driessen; Johannes Nielen; Manuel Sastry; Annelies Boonen; Frank de Vries

Disclosures

Arthritis Res Ther. 2020;22(92) 

In This Article

Methods

Data Source

Data for the present study were obtained from the Clinical Practice Research Datalink (CPRD) in the UK, previously known as the General Practice Research Database (http://www.cprd.com). CPRD contains the computerized medical records of approximately 13 million patients under the care of general practitioners (GPs) in the UK, representing 6.9% of the total UK population.[14] Practices contribute to CPRD only if their data quality meets research standards. Since 1987, data recorded in the CPRD include demographic information, prescription details, lifestyle parameters, clinical events, preventive care provided, and specialist referrals. CPRD has been extensively validated[15] and has been previously used to study gout.[16]

Study Population

We conducted a population-based case-control study (Figure 1). The cases consisted of all patients aged 40 years and older with a first diagnosis of gout during the period of valid data collection (from 1 January 1987 to 30 June 2014). Each case with gout was identified using READ codes.[17] READ codes are a set of clinical codes used in primary care in the UK for the registration of clinical diagnosis, processes of care (tests, screening, symptoms, patient administration, etc.), and medication. Each case with gout was matched by year of birth, sex, and practice to up to two randomly selected controls without a diagnosis of gout using incidence density sampling.[18] The date of the first recorded diagnosis of gout defined the index date for the cases and controls were assigned the same index date as their matched case. Cases and controls with a history of exposure to colchicine and uric acid-lowering therapy (ULT) (allopurinol, febuxostat, and/or uricosuric drugs) before the index date as well as their matched case or control were excluded.

Figure 1.

Flow chart, study population

Exposure and Potential Confounders

Clinical READ codes were used to determine OSA exposure. Cases and controls with a read code for OSA before the index date were classified as being exposed to OSA.

The following variables were considered as potential confounders and were assessed prior to the index date: smoking status, BMI, alcohol use, socioeconomic status, a history of hypertension, diabetes mellitus (as recorded by either a diagnostic code for diabetes mellitus or a history of prescription(s) for anti-diabetic treatment, British National Formulary Chapters 6.1.1 and 6.1.2), hypercholesterolemia, postmenopausal status/hysterectomy, acute myocardial infarction, stroke, or heart failure. The use of the following medication was assessed in the 6 months before the index date: thiazide diuretics, loop diuretics, beta-blockers, calcium channel blockers, angiotensin-converting enzyme inhibitors (ACE-inhibitors), angiotensin II receptor blockers (ARBs), low-dose aspirin, statins, non-insulin antidiabetic drugs (NIADDs), insulin, or benzodiazepines. In addition, the most recent eGFR before the index date was assessed. Electronic lab test data were used to extract the eGFR. Furthermore, when only serum creatinine measurements were available, these were used to estimate the eGFR by the use of the abbreviated MDRD formula (186 × (serum creatinine/88.4)−1.154 × (age)−0.203 × (0.742 if female)). In addition, we identified diagnostic codes for stages of CKD. When there were multiple records on the same day, the best eGFR was chosen. The following categories were used to stratify for renal function by eGFR: CKD 1 (eGFR > 90 ml/min), CKD 2 (eGFR 60–89 ml/min), CKD 3 (eGFR 30–59 ml/min), CKD 4 (eGFR 15–29 ml/min), and CKD 5 (< 15 ml/min).

Statistical Analysis

Conditional logistic regression was used to estimate the risk of gout associated with a diagnosis of OSA (SAS version 9.4, PHREG procedure). In the analyses, risk was expressed as odds ratios (OR) with corresponding 95% confidence intervals (CIs). Potential confounders were included in the final model if they independently changed the beta-coefficient for OSA by at least 5% or when a consensus about inclusion existed within the team of researchers, supported by clinical evidence from the literature. Missing data of confounders such as BMI, smoking status, alcohol use, and renal function were treated as separate levels using dummy variables. OSA exposure was further stratified by gender, age categories, and the presence of important confounders. Finally, we studied the effect of univariately adding the most important confounders to the main analyses as well as adjusting simultaneously for these confounders.

Comments

3090D553-9492-4563-8681-AD288FA52ACE

processing....